| 04:00 | <hober> | elisp tokenizer now passing 1281/2932 (43.7%) of tests |
| 04:00 | <hober> | most of the failures are simply that my test runner doesn't handle many of the tests right yet |
| 04:15 | <karlcow> | http://blog.jackadam.net/2010/alpha-jpegs/ |
| 04:15 | <karlcow> | I wanted a reasonably sized photographic image with a 24-bit alpha channel. So I used a JPEG for what JPEGs are good for and a PNG for what PNGs are good for… |
| 04:15 | <karlcow> | I combined them using an HTML5 canvas element and then inserted into the DOM. The results look the same as using a normal 24-bit PNG but are one-half to one-sixth the size. In one case we got a 573KB 24-bit PNG down to a 49KB JPEG with a 4KB PNG alpha-mask! |
| 08:16 | <asmodai> | lol |
| 08:16 | <asmodai> | latest Minefield crashes consistently trying to load html5test.com |
| 08:17 | <hsivonen> | asmodai: have you filed a bug with the crash reporter id already? |
| 08:17 | <asmodai> | nope, going to do it now, just wanted to try a few times to see if it was consistent |
| 08:18 | <asmodai> | hsivonen: crash reporter id, is that something different from just a bugzilla report? |
| 08:18 | <asmodai> | hsivonen: I don't even get the firefox popup |
| 08:18 | <asmodai> | cute, same happens opening the addons pane |
| 08:19 | <asmodai> | There must be some code path that's triggering NX or something similar |
| 08:19 | <hsivonen> | asmodai: If the crash reporte shows up and you submit a crash report, the next time you launch Minefield, you can navigate to about:crashes and find the report ID, so you can paste it to a bug report to allow bugzilla readers make the association between your bug report and your crash report |
| 08:20 | <asmodai> | Ah no, I get a Windows popup to Debug or Close Program |
| 08:22 | <hsivonen> | no crash for me on html5test.com on Mac |
| 08:22 | <asmodai> | Peculiar. |
| 08:22 | <asmodai> | Don't have a second box here to test with :( |
| 08:24 | <asmodai> | jeez |
| 08:24 | <asmodai> | even with disabled addons (not that I had any) and reset to all default options |
| 08:24 | <asmodai> | Really strange |
| 08:25 | <hsivonen> | if you choose to debug, do you get something that lets you get the stack trace? |
| 08:26 | <asmodai> | Nope, need to see what's up with that. Probably pointing to a previous install of Visual Studio. *checks to see if DrWatson or similar still exist in 7* |
| 08:27 | <asmodai> | hsivonen: Mmm, lets see if I can generate a .dmp file |
| 08:28 | <asmodai> | Ah |
| 08:28 | <asmodai> | event viewer has some details |
| 08:28 | <asmodai> | Faulting application name: firefox.exe, version: 2.0.0.3930, time stamp: 0x4caaf485 |
| 08:28 | <asmodai> | Faulting module name: mozjs.dll, version: 0.0.0.0, time stamp: 0x4caae92b |
| 08:28 | <asmodai> | Exception code: 0xc0000005 |
| 08:28 | <asmodai> | Fault offset: 0x00000000001ab60e |
| 08:29 | <asmodai> | same offset every single time |
| 08:29 | <asmodai> | Oh wait, the other two of the 6 have different ones. grrr. |
| 08:29 | <hsivonen> | asmodai: https://developer.mozilla.org/en/Using_the_Mozilla_symbol_server has instructions for getting symbols |
| 08:31 | <zcorpan> | annevk: missing </title> in the first example (progress events) |
| 08:50 | <asmodai> | hsivonen: it's expected that a bunch of files default to export symbols? |
| 08:53 | <hsivonen> | asmodai: I don't know |
| 08:54 | <hsivonen> | (I'm generally unhappy about builds not coming with symbols, but I realize there are download size issues.) |
| 08:55 | <asmodai> | .*** WARNING: Unable to verify checksum for C:\Program Files (x86)\Minefield\mozjs.dll |
| 08:55 | <asmodai> | *** ERROR: Symbol file could not be found. Defaulted to export symbols for C:\Program Files (x86)\Minefield\mozjs.dll - |
| 08:56 | <hsivonen> | hmm. I wonder if someone forgot to update the symbol server when mozjs.dll was split out of the libxul.dll |
| 08:56 | <hsivonen> | asmodai: we've now reached the end of my Windows debugging knowledge. Sorry. |
| 08:57 | <asmodai> | That's ok |
| 08:57 | <asmodai> | I'm following: |
| 08:57 | <asmodai> | https://developer.mozilla.org/en/How_to_get_a_stacktrace_with_WinDbg |
| 08:58 | <asmodai> | Just curious why my nightly doesn't seem to reflect the symbol server |
| 09:07 | <zcorpan> | annevk: missing </title> in the first example (progress events) |
| 09:08 | <zcorpan> | annevk: the table says 'one or more' for progress, but i could imagine zero being fired in teh case of error or abort |
| 09:09 | <annevk> | I guess |
| 09:17 | <asmodai> | aha |
| 09:17 | <asmodai> | http://symbols.mozilla.org/firefox/mozjs.pdb/4D544329A0324F3386C1DE0A369E3DB51/mozjs.pdb not found |
| 09:18 | <annevk> | I already fixed that </title> thing btw, forgot to escape the leading < |
| 09:20 | <zcorpan> | yay http://www.onenewspage.com/news/Technology/20101005/15694625/Fake-Browser-Warnings-Dupe-Users-Into-Downloading-Scareware.htm |
| 09:24 | <hsivonen> | making a distinction between what comes from the browser and what doesn't is a hard problem :-( |
| 09:25 | <asmodai> | Pink thick borders! |
| 09:55 | <asmodai> | hsivonen: got a stack trace |
| 09:56 | <asmodai> | It's a clear access violation |
| 09:56 | <asmodai> | mozjs!JS::PerfMeasurement::canMeasureSomething+0xde: |
| 09:56 | <asmodai> | 000007fe`ee30b60e ffe0 jmp rax {fffb8000`11b665b0} |
| 09:59 | <asmodai> | hsivonen: I assume the log file of the trace is better as attachment rather than inline in the bug report? |
| 10:00 | <hsivonen> | asmodai: depends on how long it is |
| 10:00 | <hsivonen> | asmodai: either way works |
| 10:01 | <hsivonen> | asmodai: I usually paste the interesting parts of the stack inline |
| 10:01 | <asmodai> | I guess that would be the exception analysis in this case |
| 10:01 | <asmodai> | k, will make sure to paste that inline and attach the full file |
| 10:03 | <micheil> | Hixie: what's with the two different sets of instructions on parsing websocket data for client and server in the spec, shouldn't both do the same thing? |
| 10:07 | <jgraham> | Ah, annevk won |
| 10:07 | <annevk> | a price? |
| 10:09 | <jgraham> | s/c/z/ ? |
| 10:10 | <asmodai> | hsivonen: Help > About was more useful when you could copy/paste the UA string :( |
| 10:13 | <zcorpan> | for those not following hybi, abarth has proposed a new handshake: http://www.ietf.org/mail-archive/web/hybi/current/msg04285.html |
| 10:13 | <micheil> | zcorpan: I'm not sure how that can actually work. |
| 10:14 | <micheil> | like, there's no indication that either wants to upgrade the connection |
| 10:15 | <jgraham> | micheil: The sent string is a unique websockets string |
| 10:16 | <micheil> | yeah, but it's just one header. |
| 10:16 | <micheil> | I think the current way is a bit better, because it uses the http Upgrade method |
| 10:16 | <micheil> | or, yeah. |
| 10:17 | <micheil> | in order to implement a server that does regular HTTP and Websockets, servers using this proposal would have to read / touch every single request |
| 10:17 | <micheil> | not just those with an upgrade. |
| 10:17 | <micheil> | (I suppose in a way that already happens though.) |
| 10:18 | <jgraham> | Well *some* part of the server has to touch every request anyway |
| 10:18 | <jgraham> | The difference is here you look for CONNECT requests to the special destination |
| 10:19 | <jgraham> | Rather than GET requests with Upgrade |
| 10:22 | <asmodai> | wth |
| 10:22 | <jgraham> | Umm, am I misunderstanding something |
| 10:23 | <jgraham> | Or is glazou proposing that we change XML parsing? |
| 10:23 | <asmodai> | so first MS installs a Bing Bar addon without my approval within Firefox and now they have a security vulnerability and they won't maintain it anymore |
| 10:23 | <asmodai> | Curious way to do that |
| 10:28 | <micheil> | jgraham, zcorpan: is this new handshaking model proposing that all data sent after handshake be encoded or encrypted too? |
| 10:30 | <jgraham> | micheil: Yes |
| 10:30 | <micheil> | isn't that what wss is for? |
| 10:30 | <jgraham> | Otherwise some protocols will just ignore all the junk they don't understand and then read the attacker-controlled bytes later |
| 10:30 | <micheil> | surely encrypting or encoding all data sent and received after handshake would slow down the rate at which data could be exchanged? |
| 10:31 | <jgraham> | No, wss is for WebSockets-over-tls |
| 10:31 | <jgraham> | That protects against people reading the network data |
| 10:31 | <micheil> | I've still got to read the document in full. |
| 10:31 | <jgraham> | This does not because the keys are exchanged in the clear |
| 10:32 | <jgraham> | It is *only* designed to prevent the attacker controlling the precise series of bytes that go over the wire |
| 10:34 | <micheil> | hmm.. |
| 10:34 | <jgraham> | It is more work for servers for sure |
| 10:35 | <asmodai> | hsivonen: https://bugzilla.mozilla.org/show_bug.cgi?id=602162 |
| 10:35 | <micheil> | the other thing to look at is just how it breaks backward compatibility |
| 10:36 | <jgraham> | micheil: In theory there is no back-compat promised at the moment |
| 10:37 | <micheil> | hmm, true |
| 10:37 | <micheil> | a bit of a pain in the ass for developers. |
| 10:38 | <jgraham> | Well yes, but presumably so is having all your email deleted due to a cross-protocol attack |
| 10:38 | <jgraham> | So there is at least some [ptential benefit |
| 10:43 | <micheil> | jgraham: do you think the websocket protocol even needs to handle extensions and such? |
| 10:44 | <jgraham> | micheil: Sure. Things like compression and multiplexing can be optional extensions |
| 10:44 | <micheil> | hmm.. |
| 10:44 | <jgraham> | So the people with difficult performance requirements can use them and the rest of us can get on with out lives |
| 10:44 | <micheil> | I think compression would be best done by end-developer |
| 10:44 | <jgraham> | *our |
| 10:44 | <jgraham> | I continue to think that every simple developer reimplementing compression in javascript is crazy |
| 10:45 | <jgraham> | *single |
| 10:45 | <micheil> | as for multiplexing, I'm not sure on the specifics of it. |
| 10:45 | <jgraham> | although simple works too |
| 10:45 | <micheil> | well, does every developer reimplement DOM methods in javascript? |
| 10:45 | <micheil> | most javascript developers are likely to use a framework that implements DOM methods, for finding nodes etc. |
| 10:46 | <jgraham> | No, DOM implements those methods |
| 10:46 | <jgraham> | in native code |
| 10:46 | <micheil> | uhh.. |
| 10:46 | <jgraham> | Developers sometimes use frameworks to provide a more convenient API |
| 10:46 | <micheil> | I'm meaning the things like what most javascript libraries do. |
| 10:47 | <jgraham> | But moving more stuff to native code made those libraries easier to implement and faster |
| 10:47 | <jgraham> | e.g. querySelectorAll |
| 10:47 | <micheil> | true, I guess. |
| 10:48 | <micheil> | so are things now heading towards per-connection extensions instead of per-packet extensions? |
| 10:48 | <jgraham> | I'm not sure |
| 10:48 | <jgraham> | With any luck people have stopped worrying about extensions for now |
| 10:48 | <micheil> | because at any rate, per-connection extensions make a lot more sense. |
| 10:49 | <jgraham> | Yes |
| 10:51 | <micheil> | so, will CONNECT 1C1BCE63-1DF8-455C-8235-08C2646A4F21.invalid:443 HTTP/1.1 |
| 10:51 | <micheil> | ever change? |
| 10:51 | <jgraham> | In theory, no |
| 10:51 | <micheil> | okay |
| 10:52 | <micheil> | so in that case you could listen for CONNECT requests where the, i guess, resource is 1C1BCE63-1DF8-455C-8235-08C2646A4F21.invalid:443 |
| 10:54 | <jgraham> | I think that is the idea |
| 10:54 | <micheil> | okay, that makes more sense now, I was reading it as that resource would change each time |
| 10:56 | <micheil> | that leads to another question: currently I've seen a lot of people doing websocket routing, where by they use the resource requested: GET <resource> HTTP/1.1 |
| 10:56 | <micheil> | to route to different channels |
| 10:58 | <micheil> | how would that be handled using this new method? |
| 11:00 | <jgraham> | You would add some mechanism for that in the post-encryption part of the handshake |
| 11:02 | <micheil> | hmm.. |
| 11:03 | <micheil> | well, that's a very common pattern, so it'd be good to see it considered. |
| 11:04 | <jgraham> | Right, abarth specifically said that he wasn't detailing the post-encryption part |
| 11:17 | <hsivonen> | asmodai: thanks for filing the crash |
| 11:18 | <hsivonen> | asmodai: yeah, it's inconvenient that the new about box doesn't contain the date for nigthlies |
| 11:28 | <Philip`> | asmodai: You can get the UA string by opening "about:" |
| 11:47 | <asmodai> | Philip`: ah |
| 12:31 | <micheil> | jgraham: did you notice a few grammatical / spelling errors in that document? some of it just doesn't seem to read right. |
| 12:32 | <asmodai> | Mmm |
| 12:32 | <asmodai> | the addon crash is different though |
| 12:34 | <micheil> | Hixie: I think there could be a security issue with the current websocket implementation of subprotocols. |
| 12:36 | <micheil> | or not. |
| 12:37 | <annevk> | you realize he's no longer editing or working on the protocol, right? |
| 12:39 | <micheil> | yeah, I thought it might've been something from the web idl side, which he is working on, iirc. |
| 12:40 | <micheil> | basically the thing consisted of: conn = new WebSocket("ws://localhost/test";, "test\r\nX-Origin: Hello"); |
| 12:41 | <micheil> | but the \r and \n aren't matched to any-char, which was an oversight on my behalf. |
| 12:47 | <zcorpan> | micheil: the latest draft of websockets api also doesn't allow space, but can still produce a space on the wire using new WebSocket(url, ['foo', 'bar']) |
| 12:47 | <micheil> | hmm |
| 12:47 | <zcorpan> | but i think that's not implemented anywhere yet |
| 12:48 | <micheil> | the array of DOMStrings seems more javascript like then a string with spaces |
| 12:48 | <micheil> | hence the reason I suggested that format. |
| 12:48 | <micheil> | iirc, ["test", "foo"].toString() -> "test foo" |
| 12:49 | <zcorpan> | no, it toString()s to test,foo |
| 12:49 | <micheil> | erm, or not. |
| 12:49 | <micheil> | yeah |
| 12:50 | <micheil> | where's the latest spec version? |
| 12:50 | <micheil> | it's not this: http://tools.ietf.org/html/draft-hixie-thewebsocketprotocol-76 |
| 12:51 | <jgraham> | micheil: Yeah there are some minor spelling / grammar errors in that document |
| 12:51 | <micheil> | nor does it appear to be http://tools.ietf.org/html/draft-ietf-hybi-thewebsocketprotocol-02 |
| 12:51 | <micheil> | jgraham: okay, hopefully I've interpreted it right. |
| 12:51 | <zcorpan> | -02 is the latest version i'm aware of |
| 12:52 | <micheil> | zcorpan: okay, well that still allows spaces in the sub-protocols by the looks |
| 12:55 | <zcorpan> | then ian fette failed to merge in the change |
| 12:55 | <micheil> | okay |
| 12:56 | <zcorpan> | http://html5.org/tools/web-apps-tracker?from=5172&to=5173 |
| 12:57 | <asmodai> | hsivonen: is there any standard way for titles/description for crashes due to executing stuff in memory marked no-execute? |
| 12:58 | <micheil> | "The subprotocol names must all be non-empty ASCII strings with no control characters and not spaces in them" |
| 12:58 | <micheil> | should be: |
| 12:58 | <micheil> | "The subprotocol names must all be non-empty ASCII strings with no control characters and no spaces in them" |
| 12:59 | <hsivonen> | asmodai: I don't know |
| 12:59 | <asmodai> | hsivonen: ok, I'll just think of something |
| 12:59 | <asmodai> | hsivonen: turns out the addon page crash is different from the html5test one I encountered -- thought at first it was related |
| 13:00 | asmodai | wonders where this instability came from so late in the beta cycle |
| 13:02 | Philip` | suspects it came from bugs |
| 13:02 | jgraham | suggests bugs in Jägermonkey in particular |
| 13:03 | <zcorpan> | micheil: file a spec bug |
| 13:03 | <micheil> | zcorpan: how? |
| 13:04 | <micheil> | zcorpan: also, do you know if the examples of data packets in the latest draft for websockets is correct? |
| 13:04 | <zcorpan> | micheil: e.g. from the comment box in http://www.whatwg.org/specs/web-apps/current-work/complete/network.html |
| 13:05 | <zcorpan> | micheil: which examples? |
| 13:05 | <hsivonen> | asmodai: radically major Jägermonkey stuff is landing |
| 13:05 | <micheil> | http://tools.ietf.org/html/draft-ietf-hybi-thewebsocketprotocol-02#section-4.6 |
| 13:06 | <zcorpan> | micheil: well i guess they are correct for -02, but browsers don't implement -02 |
| 13:07 | <micheil> | hmm.. so, that means all fields that are 0x00 or %x0 are optional? |
| 13:07 | <micheil> | zcorpan: I'm just wanting to understand where the spec is heading. |
| 13:07 | <micheil> | s/spec/protocol |
| 13:09 | <annevk> | today is a good day to answer http://www.w3.org/2002/09/wbs/40318/issue-41-objection-poll/ |
| 13:11 | <kamathln> | annevk: is this channel only meant for core whatwg people ? the page is asking for auth |
| 13:12 | <annevk> | kamathln, nope, you can probably view http://www.w3.org/2002/09/wbs/40318/issue-41-objection-poll/results |
| 13:13 | <asmodai> | hsivonen: https://bugzilla.mozilla.org/show_bug.cgi?id=602174 |
| 13:13 | <annevk> | this channel is for everyone and while somewhat focused does not have a specific topic and is most certainly without logic ;p |
| 13:13 | <asmodai> | Philip`: lol, obviously ;) |
| 13:14 | <asmodai> | hsivonen: clear enough you think? |
| 13:16 | <hsivonen> | asmodai: yeah |
| 13:16 | <hsivonen> | asmodai: might be related to GC compartments |
| 13:17 | <micheil> | zcorpan: hmm.. simon pieters is you? |
| 13:17 | <asmodai> | hsivonen: Well, I'm more than happy enough to provide more data if I am told how to get said data ;) |
| 13:19 | <zcorpan> | micheil: yes |
| 13:20 | <zcorpan> | micheil: what do you mean with optional? |
| 13:20 | <micheil> | nice to meet you then; anther irc nick to a name. |
| 13:20 | <micheil> | zcorpan: well, I was expecting that all the fields where needed |
| 13:20 | <zcorpan> | yeah sorry for the nick, i've had it for a long time :) |
| 13:20 | <micheil> | such that: 0x04 0x05 "Hello" |
| 13:20 | <micheil> | should actually be: |
| 13:21 | <micheil> | 0x00 0x00 0x00 0x00 0x04 0x00 0x05 0x00 "Hello" |
| 13:22 | <micheil> | or am I totally misunderstanding this stuff? |
| 13:22 | <zcorpan> | i think yes :) |
| 13:22 | <micheil> | which? |
| 13:22 | <zcorpan> | you're misunderstanding |
| 13:23 | <micheil> | okay |
| 13:25 | <micheil> | zcorpan: would you be able to explain how I'm misunderstanding, like, it doesn't make too much sense to me to obmit what seems like actual data |
| 13:25 | <zcorpan> | micheil: the framing in -02 is defined down to each individual bit |
| 13:27 | <zcorpan> | micheil: see "A high level overview of the framing |
| 13:27 | <zcorpan> | is given in the following figure." |
| 13:29 | <zcorpan> | so the first byte being 0x04 means that only one of the opcodes are set and the other bits are not set |
| 13:29 | <micheil> | am I right in reading that each time there is a horizontal segment to that figure that it means that it actually continues on? |
| 13:30 | <micheil> | I'm just trying to figure out how a parser may parse this new packet format |
| 13:31 | <micheil> | zcorpan: I'm guessing reading these diagrams would be something covered in an ITC / Network course at university level, right? |
| 13:32 | <zcorpan> | to find out if the text frame opcode is set, i guess you'd do something like 0x04 & first-byte |
| 13:32 | <zcorpan> | micheil: dunno :) |
| 13:33 | <micheil> | zcorpan: hmm, I'm hoping it will. |
| 13:34 | <micheil> | at any rate, I'm glad I'm re-writing node-websocket-server's parser, as it should make it easier to fork and implement this new packet format. |
| 13:35 | <zcorpan> | hmm, actually i guess 0x04 & first-byte would also be true if the opcode is e.g. 0x05 |
| 13:35 | <Philip`> | micheil: Are you intepreting the "MORE", "RSV1" etc boxes as being a whole byte? |
| 13:35 | <micheil> | Philip`: I'm not sure. |
| 13:36 | <Philip`> | (when suggesting "0x00 0x00 0x00 0x00 0x04 ...") |
| 13:36 | <Philip`> | zcorpan: You'd want 0b00001111 & first-byte |
| 13:36 | <Philip`> | (or 0x0f & first-byte) |
| 13:37 | <Philip`> | Oh, wait, I missed the point |
| 13:37 | <Philip`> | (0x0f & first-byte) == 0x04, to detect text frames |
| 13:38 | <zcorpan> | ah, yes |
| 13:38 | <micheil> | okay |
| 13:39 | <micheil> | I'm going to read more into this / experiment more. |
| 13:39 | Philip` | hopes it's 0x0f and not 0xf0 |
| 13:39 | <Philip`> | (I'm never quite certain which endianness these things assume) |
| 13:39 | <micheil> | Philip`: I'm testing this out in a v8 js console at the same time |
| 13:40 | <micheil> | > (0x0f & 0x04) == 0x04 |
| 13:40 | <micheil> | true |
| 13:40 | <Philip`> | Yeah, it's the endianness of the spec's diagram that sometimes gets confusing |
| 13:41 | <asmodai> | hsivonen: Heh, got to love those updates on my issues by this timeless fella |
| 13:41 | <micheil> | so, for binary frame, it'd be (0x0f & 0x05) == 0x05 |
| 13:42 | <micheil> | note to self: learn more about bitwise ops. |
| 13:42 | <Philip`> | Yes, or (0x0f & 0x85) == 0x05 (if the MORE flag is set), or (0x0f & 0xf5) == 0x05 etc (if the RSV flags are set too) |
| 13:43 | <micheil> | hmm.. |
| 13:43 | <zcorpan> | micheil: it helps with understanding if you write out all the bits in binary notation |
| 13:43 | <micheil> | so you'd need to parse that byte several times to determine things |
| 13:44 | <Philip`> | &ing with 0x0f is just extracting the lower 4 bits, i.e. throwing away the MORE/RSV1/2/3 ones because you don't care about them here |
| 13:44 | <micheil> | I really should've become far better at binary counting & arithmetic. |
| 13:44 | <Philip`> | and you would & with 0x80, 0x40, 0x20, 0x10 to get each of the higher 4 bits when you do care |
| 13:47 | <micheil> | hmm, okay |
| 13:48 | <micheil> | so if MORE/RSV1/2/3 were all set, then you'd get something like.. |
| 13:48 | <micheil> | 0xf4 |
| 13:48 | <hsivonen> | what time does the HTML WG f2f end on Nov 5th? |
| 13:49 | <MikeSmith> | hsivonen: probably 5pm I guess |
| 13:49 | <hsivonen> | MikeSmith: thanks |
| 13:50 | <Philip`> | micheil: Yes |
| 13:50 | <micheil> | okay, cool. |
| 13:51 | <hsivonen> | I guess the practical thing to do then is traveling back on Saturday |
| 13:51 | <micheil> | Philip` & zcorpan: thanks for having the patience to explain things, it's really appreciated. :) |
| 14:48 | <asmodai> | hsivonen: setting javascript.options.methodjit.content=false solves the issues on both my tickets |
| 14:59 | <hsivonen> | asmodai: clearly, the bugs belong in the JS engine component, then |
| 15:01 | <asmodai> | hsivonen: I concur. |
| 15:24 | karlcow | has difficulties to see what would be the use case for PUT with forms. |
| 15:25 | <karlcow> | searching through the Web to see if some people have use cases |
| 15:26 | <karlcow> | there is also the question: Are there CMS, Web apps using it at all? |
| 15:26 | <karlcow> | I see for example http://geekswithblogs.net/narent/archive/2008/01/04/118224.aspx |
| 15:34 | <henrikbjorn> | karlcow: PUT = updating a object |
| 15:35 | <henrikbjorn> | POST = create a new one PUT = update DELETE = remove GET = retrieve |
| 15:35 | <henrikbjorn> | basic REST >P |
| 15:36 | <zcorpan> | POST handles the update and delete use cases in practice |
| 15:36 | <karlcow> | henrikbjorn: I know my HTTP… ;) what I meant is the use case for HTML FORMS |
| 15:37 | <karlcow> | authoring tools can send HTTP PUT without html forms. so the question is really about the use case for html forms |
| 15:39 | <annevk> | you realize this feature got killed right |
| 15:39 | <annevk> | ? |
| 15:42 | <karlcow> | annevk: yes precisely. I have seen some REST fans argueing, and being a REST fans too, I was trying to understand why. And sincerely I do not see a reason to maintain it. Basically I want to understand which use cases it *might* forbid. |
| 15:42 | <micheil> | is there any reason why a lot of stuff in the websocket ietf document is "something may do something" rather then "something should do something" |
| 15:43 | <micheil> | zcorpan: on that, there is a practice in rails to use a _method hidden field |
| 15:44 | <micheil> | or a methodOverride as they call it, iirc, which allows them to send PUT / DELETE requests in browsers that don't handle PUT/DELETE in forms |
| 15:44 | <micheil> | it's a hack to achieve REST |
| 15:44 | <zcorpan> | karlcow: if PUT and DELETE were restricted to same-origin submission, then using PUT or DELETE instead of POST would enable a way to protect against CSRF |
| 15:45 | Philip` | notes that REST is a conceptual architecure, not a collection of specific HTTP-level implementation details |
| 15:45 | <Philip`> | but people sometimes seem to get fixated on the latter |
| 15:45 | <zcorpan> | micheil: so the form is actually sent using POST? |
| 15:45 | <micheil> | zcorpan: yeah |
| 15:46 | <micheil> | but the server sees that the POST data includes a methodOverride field, and changes the interpreted method to be that method. |
| 15:46 | <micheil> | which then restarts the routing process. |
| 15:46 | <jgraham> | zcorpan: You could also make POST2 |
| 15:46 | <zcorpan> | jgraham: yes |
| 15:46 | <jgraham> | Since often POST semantics are what you want |
| 15:47 | <micheil> | like, you make a PUT request to a rails app, and it has the same effect as making a POST with a methodOverride set to PUT |
| 15:50 | <mamund> | annevk: it's not clear to me why browsers should only support idempotent writes (PUT, DELETE) via scripting (XmlHttpRequest), but not via direct markup (e.g. FORM method="put"). |
| 15:51 | <karlcow> | mamund: what would it solve? |
| 15:51 | <karlcow> | FORM method="put" |
| 15:51 | <mamund> | the same thing the RoR "work-around" solves |
| 15:52 | <mamund> | idempotent writes are important for dist. networ, apps. |
| 15:52 | <AryehGregor> | How can you ever guarantee that a write is idempotent? What if someone else wrote to the same resource in between? |
| 15:52 | mr_lou | thinks the <audio> tag of HTML5 ought to support MIDI files |
| 15:53 | <karlcow> | when using HTTP PUT, you know and have control on the URI you want to PUT on. Does that mean the UI cater for this? |
| 15:53 | <AryehGregor> | Likewise, for DELETE, what if someone created a new version of the resource in between? |
| 15:53 | <annevk> | mr_lou, does in Opera, iirc |
| 15:53 | <mamund> | AryehGregor: that's what ETags are for. |
| 15:53 | <AryehGregor> | mamund, and can ETags actually be provided in HTML forms? |
| 15:53 | <AryehGregor> | In the manner necessary for idempotency to hold? |
| 15:53 | <zcorpan> | annevk: not on mac and windows i don't think |
| 15:53 | <AryehGregor> | mr_lou, speak to browser vendors about it. HTML5 doesn't mandate specific media formats. |
| 15:54 | <mamund> | karlcow: the URI need not be created by the client, but supplied by the server. PUT for update is the best example |
| 15:54 | <karlcow> | AryehGregor: http://www.w3.org/1999/04/Editing/ |
| 15:54 | <mr_lou> | AryehGregor, roger |
| 15:54 | <mamund> | AryehGregor: ETags are supplied for all GETs, why not PUTs and DELETEs? |
| 15:55 | <mamund> | Amaya browser has supported PUT/DELETE w/ ETags for quite some time. |
| 15:55 | <AryehGregor> | This all seems very complicated, given that in practice the problem is typically solved at levels above HTTP anyway. HTTP doesn't seem like the right level to solve the problem at all, it gives very coarse-grained control. |
| 15:55 | <karlcow> | mamund: yes I kind of know for Amaya (given my past), but it is not using forms to do that. |
| 15:56 | <AryehGregor> | You can do all this stuff easily at the level of the web application, without having to try persuading your web server to cooperate. |
| 15:56 | <mamund> | karlcow: yes, good point |
| 15:56 | AryehGregor | shrugs |
| 15:56 | <karlcow> | we are mixing two discussions here. HTTP PUT (that I love), and HTTP PUT in forms |
| 15:57 | <mamund> | karlcow: is PUT w/ FORMS a technical issue or what? |
| 15:58 | <mamund> | i use a scripting workaround to support FORM method="put" quite often w/o trouble. |
| 15:59 | <karlcow> | mamund: http://www.w3.org/Bugs/Public/show_bug.cgi?id=10671 |
| 15:59 | <mamund> | karlcow: yep, i've seen it. |
| 16:00 | <mamund> | karlcow: "payload control" is an odd reference. (re: PUT) |
| 16:01 | <mamund> | as to the reasoning for DELETE, currently POST has 200 & 204 as potential response codes already. |
| 16:02 | karlcow | would love to read a user scenario of a real PUT and DELETE done with forms. |
| 16:03 | <mamund> | karlcow: would love to read a user scenario for a real POST done w/ forms<g>. |
| 16:03 | <karlcow> | hehe |
| 16:03 | <mamund> | srsly: what user scenario would be compelling? |
| 16:04 | <mamund> | surely the thousands of cases where XmlHttpRequest issues PUT & DELETE carries some weight, no? |
| 16:04 | jgraham | wonders how one answers that sort of question |
| 16:04 | <mamund> | is it really the "FORM" use that is unappealing here? |
| 16:04 | <jgraham> | "give me a use case" "not until you tell me which use case is compelling" |
| 16:04 | <mamund> | jgraham: LOL |
| 16:06 | <mamund> | i wonder if this would go differently if someone proposed a new element <put ... /> instead of a new enumeration for FORM method="..." |
| 16:07 | <Philip`> | If someone proposed that, people would probably respond "how does that even make sense?" |
| 16:07 | <jgraham> | mamund: "show me the use case" is not intended as a smack-down |
| 16:08 | <mamund> | jgraham: i understand. np. |
| 16:08 | <jgraham> | It's supposed to be a way of determining if there is a real need that is not being met |
| 16:08 | <jgraham> | A surprising number of "needs" are actually imaginary |
| 16:08 | <mamund> | i think the RoR workaround is an example of a common use case, is it not? |
| 16:09 | <karlcow> | because I started the discussion here about it. My "show me the use case" means I really want to understand what are the cases for method="put". Maybe there is a good use case, but I still do not see one. |
| 16:11 | <mamund> | karlcow: i use PUT to support sending updates of existing resources in an idempotent way (using ETags for concurrency) |
| 16:11 | <karlcow> | with a web form? |
| 16:11 | <mamund> | w/o PUT support for unscripted browsers, i need to mimic the same features by using POST w/ an ETag hash in the FORM. |
| 16:12 | <mamund> | karlcow: yes, with a FROM |
| 16:12 | <mamund> | s/FROM/FORM |
| 16:12 | <karlcow> | could you explain it in details in the Bugzilla? |
| 16:12 | <mamund> | i'll add a comment there, yes. |
| 16:13 | <karlcow> | thanks |
| 16:40 | <karlcow> | wiiiizzz http://benmetcalfe.com/blog/2010/10/the-ly-domain-space-to-be-considered-unsafe/ |
| 16:44 | <paul_irish> | holon lemme one up you, karlcow |
| 16:44 | <paul_irish> | http://tech.slashdot.org/story/10/10/06/1440246/W3C-Says-Dont-Use-HTML5-Yet |
| 16:45 | <karlcow> | paul_irish: why me? :) |
| 16:47 | <AryehGregor> | karlcow, so someone figured out that ccTLDs are actually run by individual countries? Gee, who'd have thought? |
| 16:47 | <AryehGregor> | I thought they were just like .com except that you can use them to make clever-looking domain names like "bit.ly"! |
| 16:47 | <AryehGregor> | I'm sure a lot of people do think that, in fact, but . . . they're wrong. No kidding you don't want to violate Libyan law on a .ly domain name. |
| 16:49 | <AryehGregor> | "Meanwhile, HTML5 is headed toward final approval in two to three years." hahahahaha right. |
| 16:49 | <AryehGregor> | CR, yeah, as in "deliberately obsolescent snapshot of a feature set no one cares about or targets anymore". |
| 16:50 | <AryehGregor> | (I think they just misunderstood plh, though, I doubt he'd have called CR "final approval".) |
| 16:51 | <AryehGregor> | What a stupid article. |
| 16:51 | <hsivonen> | at least the comments on Slashdot are pretty good this time |
| 16:52 | <karlcow> | Paul Krill seems to have something against it http://www.infoworld.com/d/developer-world/microsoft-silverlight-capabilities-exceed-those-html5-329 |
| 16:52 | <AryehGregor> | (I mean the InfoWorld one about HTML5 cited on Slashdot. Although the one whining about registering a domain name under a TLD controlled by a corrupt third-world dictatorship and then actually having to deal with a corrupt third-world dictatorship, that one's pretty stupid too.) |
| 16:53 | <paul_irish> | AryehGregor: yah plh is doing a bunch of html5 evangelism himself.. just spoke at openvideoconf about html5 video and svg.. so i dont think this is what he intended. |
| 16:55 | <karlcow> | and I do not think he actually said this ""We're not going to retire Flash anytime soon," Le Hegaret said. " |
| 17:10 | <AryehGregor> | A surprising number of people on Slashdot seem to not understand why standardization takes so long. |
| 17:10 | <AryehGregor> | Maybe someone should write up an article explaining, and get it Slashdotted. |
| 17:11 | <TabAtkins> | Sure, I'll spin up a task force to write it. |
| 17:11 | <KaOSoFt> | Perhaps that people still want tag soups, blink, and that... |
| 17:12 | <jgraham> | Can I be on the section-heading sub task force |
| 17:12 | <jgraham> | Or is it sub-committee, I forget? |
| 17:12 | <TabAtkins> | Please, there will be plenty of sub task forces for everyone. |
| 17:13 | <Philip`> | How can you even begin to consider the structure of the task force before having a conference call to discuss it? |
| 17:14 | <zcorpan> | Philip`: when's the call? have you written the agenda? |
| 17:14 | <jgraham> | Strictly we should probably set up an incubator group to assess the need for a task force in this area |
| 17:15 | <Philip`> | I don't know when - we'll need a poll to work out when the most important people are available |
| 17:15 | <jgraham> | If the incubator group produces anything we can work out a charter for the task force |
| 17:15 | <jgraham> | Mandatory telecons can go in the charter |
| 17:16 | <zcorpan> | Philip`: a poll? maybe we need a new protocol to do the poll properly |
| 17:17 | <Philip`> | That would be outside our jurisdiction so we'd have to outsource it to the IETF |
| 17:18 | <micheil> | man... >_> |
| 17:20 | <tw2113> | woman... |
| 17:20 | <tw2113> | :D |
| 17:20 | <AryehGregor> | In the computer lab at school the other day, I heard two guys discussing HTML5. |
| 17:20 | <variable> | AryehGregor, did they know who you were ;-) |
| 17:20 | <AryehGregor> | One guy was showing the other guy something, and the other guy asked what it was written in, and the first guy said something like "Pure HTML5, it's really powerful!" |
| 17:20 | <karlcow> | AryehGregor: http://www.w3.org/QA/2008/11/w3c-open-web-standards |
| 17:20 | <AryehGregor> | I tried sneaking a look, but they saw me and I didn't want to weird them out. |
| 17:21 | <AryehGregor> | It seemed to be some trivial thing with balls bouncing around or whatnot. |
| 17:21 | <tw2113> | oh just weird people out, it's worth it and they need their feather ruffled |
| 17:21 | <Philip`> | Like Google's "HTML5" thing made with coloured <div>s? |
| 17:21 | <tw2113> | it's a spice of life |
| 17:22 | <AryehGregor> | variable, I could have introduced myself, but it would have been a bit weird. |
| 17:22 | <AryehGregor> | Philip`, yeah, it looked like that. |
| 17:22 | <AryehGregor> | But I'm guessing it was done with canvas. |
| 17:22 | <tw2113> | that ball thing from google was all javascript |
| 17:25 | <AryehGregor> | Well, JavaScript is part of HTML5, effectively. |
| 17:25 | <AryehGregor> | Some HTML5 stuff is purely JS APIs. |
| 17:25 | <AryehGregor> | Like canvas. Not too useful without JS. |
| 17:26 | <jgraham> | Some of DOM is part of HTML5 |
| 17:26 | <jgraham> | Javascript isn't |
| 17:26 | <jgraham> | If we are being pedantic |
| 17:26 | jgraham | doesn't really see the point of being pedantic on this issue though |
| 17:26 | <variable> | HTML5 is an evangelicalism term |
| 17:26 | <variable> | like AJAX years ago :-| |
| 17:27 | <tw2113> | just remember...html5 canvas needs js more than js needs anything html5 |
| 17:28 | <loucapo> | hey everyone |
| 17:29 | <loucapo> | i know there is a Firefox bug related to xmlhttplevel 2 and status codes |
| 17:30 | <loucapo> | another issue i recently ran into relates to xmlhttp posts to a pwd protected url. this works ok in chrome/safari but firefox is failing with the annoying status=0 ssue |
| 17:30 | <loucapo> | anyone know if thats a known bug? |
| 17:36 | <loucapo> | here is the bug i was referring to |
| 17:36 | <loucapo> | https://bugzilla.mozilla.org/show_bug.cgi?id=597301 |
| 18:31 | <TabAtkins> | Recursive reference: https://twitter.com/miksago/status/26566110962 |
| 18:31 | <micheil> | hmm? |
| 18:32 | <TabAtkins> | Oh! I didn't even realize it was you that posted it. ^_^ |
| 18:32 | <micheil> | haha. |
| 19:30 | <jgraham> | TabAtkins: I can't believe I saw that on WHATWG IRC! |
| 19:33 | <jgraham> | In unrelated news, today's xkcd is cool |
| 19:34 | <TabAtkins> | It is! |
| 19:39 | <TabAtkins> | jgraham: Where can I get info about the html test harness? |
| 19:40 | <jgraham> | TabAtkins: Info of waht sort? |
| 19:40 | <TabAtkins> | Sufficient information to write a test using it. |
| 19:40 | <jgraham> | Hmm, the documentation is not great |
| 19:41 | <jgraham> | I can fix that now I guess |
| 19:41 | <TabAtkins> | That'd be cool. Pilgrim is asking me about it, and apparently doesn't want to start irc on his mac. |
| 19:59 | <hober> | IE9 will be shipping with the HTML5 parser, right? While I'm asking, have either Safari or Chrome shipped the parser yet? |
| 20:11 | <MikeSmith> | hober: IE9 parser is not conformant |
| 20:11 | <MikeSmith> | unfortunately |
| 20:12 | <MikeSmith> | and the WebKit HTML5 parser has not shipped in normal Safari or Chrome yet, afaik |
| 20:12 | <MikeSmith> | I think it's in the current Chrome betas though |
| 20:12 | <MikeSmith> | or I would sure think it would be by now |
| 20:13 | <MikeSmith> | Chrome 7 beta |
| 20:13 | <TabAtkins> | I *think* it is, but I'd have to ask to make sure. |
| 20:13 | <TabAtkins> | Or you can hand me a testcase that'll only pass in an HTML5-compliant parser, and I'll check what I have. |
| 20:31 | <rubys1> | hober: ping? |
| 20:32 | <hober> | rubys1: yo |
| 20:33 | <rubys> | Saw the discussion on www-archive re libxml2... |
| 20:33 | <rubys> | I think the key is to get libxml2 to pass the same tests as Henri's parser does |
| 20:33 | <hober> | oh, if the libxml2 guys want to impl the html5 parsing algorithm, i'm all for it |
| 20:33 | <hober> | that would be awesome |
| 20:34 | <rubys> | that's the key IF. Any evidence that that is the case? |
| 20:35 | <hober> | I have no evidence either way |
| 20:36 | <hober> | this thread came out of the nodejs list, on which I was advocating using Aria's JS impl of the HTML5 parser and not the libxml2 HTML parser |
| 20:40 | <rubys> | My guess is that Aria is more compliant but libxml2 is much faster. |
| 20:40 | <rubys> | quick scan, and I see a thread from 2008 about libxml2 implementing html5 that seems to have died... |
| 20:42 | <hober> | *nod*. for my needs, correct trumps fast as I'm using the parser in the publication workflow of what ends up being a pile of static assets served by apache |
| 20:42 | <hober> | scales nicely :) |
| 20:43 | <rubys> | libxml2 is used widely; I once tried to help maintain a Ruby port of the html5 library, but have slipped back to using libxml2. |
| 20:43 | <rubys> | speed wasn't the issue for me... support was. |
| 20:52 | <jarib> | rubys: is there a C port though? that would make it easy to wrap in whatever language and keep decent speed |
| 20:53 | <rubys> | hsivonen's has code that converts his java parser into C and then includes that into firefox; the problem is that that code presumes that you have mozilla's runtime library |
| 20:54 | <jarib> | right |
| 20:54 | <rubys> | I think the right thing to do is to pelt the libxml2 guys with bug reports :-P |
| 20:54 | <rubys> | Anybody remember where the test suite is for html5lib? (I used to know...) |
| 20:54 | jarib | doesn't |
| 20:55 | <hober> | hg clone https://html5lib.googlecode.com/hg/ html5lib |
| 20:55 | <hober> | then it's html5lib/testdata |
| 20:55 | <rubys> | oh, I thought it was a separate project... |
| 20:55 | <hober> | oh, could be. I've been pulling it out of html5lib for the elisp tokenizer |
| 20:56 | <MikeSmith> | I would doubt that the libxml2 HTML parser would ever pass the html5lib tests without pretty much being rewritten |
| 20:59 | <jgraham> | It isn't a seperate project |
| 21:18 | <mamund> | jgraham: posted my comments on PUT/DELETE into bugzilla. hopefully they make sense/are helpful. i'll be avail for follow up. thanks. |
| 22:02 | <hober> | Like annevk said, today is a good day to answer http://www.w3.org/2002/09/wbs/40318/issue-41-objection-poll/ |
| 22:14 | <TabAtkins> | hober: Aryeh and Henri stated everything that I would have said. |
| 22:15 | <Dashiva> | Can someone link the results page, I don't remember how to transform the URL |
| 22:15 | <TabAtkins> | http://www.w3.org/2002/09/wbs/40318/issue-41-objection-poll/results |
| 22:15 | <Dashiva> | Thanks |
| 22:16 | <Dashiva> | Oh, you just append results? |
| 22:16 | <TabAtkins> | Yeah. |
| 22:30 | <rubys> | mamund: what's the bug number? |
| 22:30 | <mamund> | hmmm |
| 22:30 | <mamund> | rubys: http://www.w3.org/Bugs/Public/show_bug.cgi?id=10671 |
| 22:32 | <MikeSmith> | in natural languages, what is the difference between lexical and syntactic? |
| 22:32 | <MikeSmith> | or in the study of linguistics |
| 22:32 | <MikeSmith> | is there a difference? |
| 22:32 | <hober> | MikeSmith: asking my wife (a computational linguist) |
| 22:33 | <rubys> | mamund: POST responses are NOT cacheable. |
| 22:33 | <mamund> | if caching directives are supplied, they can be honored, right? |
| 22:34 | <rubys> | ah, there is an 'unless' in there. |
| 22:34 | <MikeSmith> | hober: maybe syntactic = about sentences |
| 22:34 | <mamund> | yeah, not common, but possible |
| 22:34 | <MikeSmith> | or grammatically complete utterances |
| 22:34 | <MikeSmith> | whatever the general term is for those |
| 22:34 | <hober> | MikeSmith: "Lexical = to do with words \n the lexicon \n syntax = how the words relate to each other to form coherent sentences" |
| 22:34 | <hober> | yeah, you got it |
| 22:34 | <MikeSmith> | aha |
| 22:35 | <MikeSmith> | hober: thanks much |
| 22:35 | <jgraham> | My g/f said the same |
| 22:35 | <MikeSmith> | jgraham: thanks |
| 22:35 | <jgraham> | and she is a linguist by training |
| 22:35 | <MikeSmith> | ok |
| 22:36 | <jgraham> | MikeSmith: Why were you wondering? |
| 22:36 | <hober> | re: web geekery & linguistics, my wife's blog: http://snowclones.org/ |
| 22:37 | <MikeSmith> | jgraham: I was reading a text that describes V-V compounds |
| 22:37 | <MikeSmith> | compound verbs |
| 22:37 | <MikeSmith> | Japanese compound verbs |
| 22:38 | <MikeSmith> | and it says there are two types of Japanese V-V compounds |
| 22:38 | <MikeSmith> | one is lexical V-V compounds |
| 22:38 | <MikeSmith> | which is the normal kind that most such verbs fall into |
| 22:38 | <MikeSmith> | the second time is syntactic V-V compounds |
| 22:39 | <MikeSmith> | in which the second verb in the pair is a verb that relates to beginning, ending, or continuing something |
| 22:39 | <MikeSmith> | which verbs are apparently called "aspectual" verbs |
| 22:42 | <jgraham> | As in http://en.wikipedia.org/wiki/Grammatical_aspect I assume |
| 22:56 | <MikeSmith> | jgraham: yeah, seems so |
| 23:17 | <jgraham> | TabAtkins: Pushed some very incomplete documentation to testharness.js |
| 23:19 | <ojan> | does anyone know which url is the right one for the new web dom core spec? i have like 5 different versions in various emails. not sure which is the one to read and comment on. |
| 23:20 | <ojan> | gsnedders: ^^^ maybe you know? your name is on one of the specs i'm looking at. :) |